This should be an EasyFix for someone with a little dns experience. We have many web applications accessing db01 and db02 for their data. This change is to accomplish two things:

1. Make the naming of database servers be by function, rather than a generic "db". This will make it easier to see what needs to move later.
2. Lay the groundwork for splitting the databases to different hardware depending on their function.

What we need is for someone to edit the DNS entries for db01(postgres) and db05(mysql) (In production. In stg, they're both on db02) to add CNAMES for the apps that are on there. At the same time, they need to update the application's config files to point at the new DNS name for the db server. As an example,

PackageDB accesses the postgres db on db01 for it's data. So we need to add a CNAME for db-pkgdb that points to the db01 box. The pkgdb config files in puppet/modules/fedora-packagedb/templates/pkgdb-prod.cfg.erb would get updated at the same time so that the dburi line would point to:
{{{
-sqlalchemy.dburi="postgres://pkgdbadmin:<%= pkgdbPassword %>@db01/pkgdb"
+sqlalchemy.dburi="postgres://pkgdbadmin:<%= pkgdbPassword %>@db-pkgdb/pkgdb"
}}}

Notes:
* When you switch FAS, be sure to grep through puppet for other places that directly connect to the database. I think that the only other place is with http basic auth (for instance, trac on fedorahosted) which uses mod_auth_postgres.
* Naming convention for these will be db-APPNAME (no need for numbers as we're unlikely to ever have multiple db masters for an app).
* If we implement slave db servers, they'll get the name db-APPNAME-slave.
* If we have to replace a db box, we name it something temporary like db-APPNAME-new. When we're ready to switch, we take an outage (necessary to sync final copies of data anyhow) and change the names in DNS.